ERROR GROWTH IN NUMERICAL PREDICTION AND ATMOSPHERIC PREDICTABILITY

The article is to report some results of numerical experiments on the error growth and the atmospheric predictability Experiments with two-level global baroclinic primitive equation spectral model have main results as follows.The magnitude of initial errors directly affects the error growth,but its distribution form has little effect on the growth.The loss of predictability resulting from small-scale error is much greater than that from large-scale error.The small-scale error rapidly grows and is transferred to the large-scaleerror by interaction between different scale waves,which stimulates the growth of error for the whole system Orographic forcing restrains planetary-scale error (wavenumbers 0-3) but enhances the small-scale error (wavenumbers 8 or greater).Hence,orographic effects on the error growth closely depend on the characteristic scale of initial errors,and there may be a critical wavenumber between 4 and 7.The error growth is greater in Northern Hemisphere than in Southern Hemisphere if initial errors are the same.In the end we give some discussions about model,initialization scheme,etc.,to improve model prediction.